Structural determination of the large photosystem II–light-harvesting complex II supercomplex of Chlamydomonas reinhardtii using nonionic amphipol

In photosynthetic organisms, photosystem II (PSII) is a large membrane protein complex, consisting of a pair of core complexes surrounded by an array of variable numbers of light-harvesting complex (LHC) II proteins. Previously reported structures of the PSII–LHCII supercomplex of the green alga Chlamydomonas reinhardtii exhibit significant structural heterogeneity, but recently improved purification methods employing ionic amphipol A8-35 have enhanced supercomplex stability, providing opportunities for determining a more intact structure. Herein, we present a 5.8 Å cryo-EM map of the C. reinhardtii PSII–LHCII supercomplex containing six LHCII trimers (C2S2M2L2). Utilizing a newly developed nonionic amphipol–based purification and stabilizing method, we purified the largest photosynthetic supercomplex to the highest percentage of the intact configuration reported to date. We found that the interprotein distances within the light-harvesting complex array in the green algal photosystem are larger than those previously observed in higher plants, indicating that the potential route of energy transfer in the PSII–LHCII supercomplex in green algae may be altered. Interestingly, we also observed an asymmetric PSII–LHCII supercomplex structure comprising C2S2M1L1 in the same sample. Moreover, we found a new density adjacent to the PSII core complex, attributable to a single-transmembrane helix. It was previously unreported in the cryo-EM maps of PSII–LHCII supercomplexes from land plants.

ated with inherent resolution limitations, to cryo-EM, which provides higher resolution and more biologically relevant conditions.
Cryo-EM (13) has progressed tremendously in the last two decades (14) thanks to advances such as direct electron detectors, automated data collection, and phase-contrast plates and is now able to challenge X-ray crystallography for obtainable resolution (15) while providing advantages through the ability to elucidate multiple conformations of a protein or complex simultaneously (16). Thus, cryo-EM is poised to provide sweeping new insights into photosynthetic complexes in an increasingly biologically relevant context across the range of photosynthetic organisms. Cryo-EM studies of the PSII-LHCII supercomplex have thus far focused on land plants, with reports from Arabidopsis thaliana (17), Spinacia oleracea (18), and Pisum sativum (3) detailing the structure of the C 2 S 2 and C 2 S 2 M 2 forms of PSII-LHCII supercomplex between 5.3 and 2.7 Å, whereas Albanese et al. (19) have reported a lower-resolution map from Pisum showing interaction between stacked complexes. Based on the cryo-EM map, Wei et al. (18) proposed the energy transfer route within a C 2 S 2 form of PSII-LHCII supercomplex in land plants, describing potential routes of excitation transfer from the LHCII trimer via CP29 (LHCB4) into one core complex and via CP26 (LHCB5) into the opposite core. This was expanded upon by Su et al. (3) to include the M-trimers and CP24 (LHCB6), where excitation from the M-trimer can transfer to the S-trimer or route directly through CP29.
On the other hand, C. reinhardtii does not possess CP24 (LHCB6) (6), but the PSII-LHCII supercomplex harbors additional LHCII trimers, L-trimers (10). The L-trimers have been suggested to act as stabilizers for the complexation of the major LHCII components of the PSII-LHCII supercomplex (10,20) while simultaneously increasing the available cofactors for energy absorption and ultimately photochemical energy available to the alga.
According to previous reports, the C. reinhardtii PSII-LHCII supercomplex exhibited significant heterogeneity when studied by detergent solubilization and negative stain TEM (10). The recent improvement in purification methodology employing ionic amphipol A8-35 has enhanced the stability of the supercomplex (11), thus providing us with the possibility of determining its more intact structure. Herein, we present the cryo-EM structure of the C 2 S 2 M 2 L 2 form of PSII-LHCII supercomplex in C. reinhardtii at a resolution of 5.8 Å, when purified according to the newly developed protocol employing nonionic amphipol and using the new processing suite of computational imaging system for TEM (cisTEM) (21).

Purification of PSII-LHCII supercomplex using NAPol
Previously, we developed a procedure for purifying the PSII-LHCII supercomplex of C. reinhardtii employing an ionic amphipol, A8-35 (11). Although the obtained supercomplex showed little LHCII dissociation even 4 days after purification, to maintain the oxygen-evolving activity, betaine addition was required (11). Because the decrease of oxygen-evolving activity was correlated with dissociation of the membrane-extrinsic polypeptides, the ionic nature of A8-35 was likely the cause of this instability. To overcome this problem, we introduced nonionic amphipol NAPol (22) in place of ionic amphipol A8-35 in this study. Fig. 1A shows a comparison between n-dodecyl-␣-D-maltoside (␣-DDM)-solubilized PSII-LHCII supercomplexes from C. reinhardtii (␣-DDM) according to the previous method (10), and those after NAPol substitution (NAPol), according to the newly developed method. The sucrose density gradient (SDG)

Structure of C. reinhardtii C 2 S 2 M 2 L 2 -type PSII supercomplex
profile of the NAPol sample was shifted to higher density, indicating an overall increase in the molecular weight of all isolated components from the thylakoid membrane. SDS-PAGE showed that the polypeptides in the supercomplexes isolated by the two methods were essentially identical, with moderate increases in intensity for the bands of LHCIIs, labeled types I, III, and IV and PsbQ, as well as slight comigration contamination of AtpA/B in the high-molecular-weight region (Fig. 1A). The NAPol-stabilized sample had a low-level PSI contamination as in the ␣-DDM-solubilized sample (22-24% of the PSI amount contained in the thylakoid membrane when normalized by D1 protein) (Fig. S1), which would not be a problem for single-particle analysis.
The NAPol-stabilized PSII-LHCII supercomplex showed a slightly decreased specific oxygen-evolving activity when compared with the ␣-DDM-solubilized PSII-LHCII supercomplex (Table 1), similarly to another amphipol-stabilized PSII-LHCII supercomplex using ionic amphipol A8-35 (11). These slight reductions of specific activities observed in the amphipol preparations were either due to the increased number of LHCIIs per reaction center or due to the hindrance of the enzyme dynamics by amphipol, which is called the "Gulliver effect" (23). Because the NAPol-stabilized sample showed a lower Chl a/b ratio than the ␣-DDM-solubilized sample (Table 1), we favored the former hypothesis. We also quantified cytochrome b 559 in the NAPol sample: Assuming that two units of cytochrome b 559 were in a supercomplex, the PSII-LHCII supercomplex was quantified to bind 354 Chls, suggesting that ϳ5.4 LHCII trimers were associated with a supercomplex. As already shown in the A8-35-stabilized preparation, these amphipol-stabilized PSII-LHCII supercomplexes are more stable than the conventional detergent-solubilized preparations in terms of association of the peripheral LHCII (11). Moreover, the oxygen-evolving activity of the NAPol-stabilized PSII-LHCII supercomplex after 3 days of incubation at 4°C was 198 mol O 2 (mg Chl) Ϫ1 h Ϫ1 , which corresponded to 93% of its initial activity (Table 1). We concluded the integrity of the PSII-LHCII supercomplexes were kept intact during the NAPol-based method. The NAPolstabilized PSII-LHCII supercomplex was then subjected to the following cryo-EM single-particle analysis.

Single-particle analysis of PSII-LHCII supercomplex
The high-resolution images of the ice-embedded NAPol-solubilized PSII-LHCII supercomplex of C. reinhardtii were collected using a 200 kV electron microscope with a cryo-specimen holder, energy filter, and direct detector (see details under "Experimental procedures"). After correcting motion in the frames, 126,744 particles were picked automatically from 498 micrographs ( Fig. 1B and Fig. S2, A and B). "Bias-free" particle picking via Gaussian "blob" (21) is fairly poor at selecting "side views" of the PSII-LHCII or PSI-LHCI complexes. This can be attributed to the nature of the photosystem complexes, which are large and flat, leading to very different 2D projection size and contrast between side and top views (Figs. S2-S4). This is exacerbated by NAPol stabilization, which increases the overall size of the particle from 31 nm (longest dimension) to 36 nm with the thick-surrounding NAPol ring (Fig. 2). Separating the narrow, high-contrast side views from the broader, lower-contrast top views is not difficult; however, cleanly separating the subtypes of top view (C 2 S 2 M 2 L 2 and C 2 S 2 M 1 L 1 ) requires finer grain classification. Similarly, side views can also be difficult to cleanly separate; as a result, the different types of view were separated into different groups after the initial classification (Figs. S2, C-E, S3, and S4).
The finally selected 14,668 particles were used for 3D reconstruction of the full PSII-LHCII complex (C 2 S 2 M 2 L 2 ) (Fig.  S2H), although the angular distribution was slightly dominated in the top views (Fig. S2I). The resolution was estimated as 5.8 Å with gold standard fourier shell correlation (Fig. S2J) by imposing C2 symmetry after several local alignments with a low-pass filtered mask. Fig. 2 (A-C) shows the 5.8 Å reconstruction of C. reinhardtii PSII-LHCII supercomplex with six LHCII trimers bound, viewed from the membrane-perpendicular direction (Fig. 2, A and D) and the membrane parallel directions (Fig. 2, B, C, E, and F). The atomic models of pea PSII core complex and LHCIIs (PDB code PDB code 5XNL; 3) are individually fitted into each density (Fig. 2, D-F) to evaluate the cryo-EM map using "Fit in map" in UCSF Chimera (24). The NAPol stabilized supercomplex shows the largest dimensions of 31 ϫ 22 nm reported so far (Fig. 2, A and D). The transmembrane edge of the supercomplex was fully surrounded by a NAPol ring of ϳ2.5 nm in width. Fig. S3 shows a local resolution map produced by ResMap (25) of the density map shown in Fig. 2 colored by estimated resolution. This allows visualization of the low-resolution nature of the stabilizing NAPol, particularly in the gap between the extrinsic oxygen-evolving complex (OEC) region (Fig. S3A), which exhibits extra density. Fig. S3 (B, D, and F) shows sliced through equivalents of Fig. S3 (A, C, and E) to permit visualization of the transmembrane helices and internal density, which is much higher resolution than the NAPol envelope. This supports the overall flexibility of the supercomplex and the Gulliver effect envelope nature of amphipol. The slice views demonstrate the structural stability of the core region of the complex with clear ␣-helices, which achieves ϳ5 Å resolution, whereas extrinsic regions such as M-and L-trimers resolve to ϳ7 Å and the NAPol envelope at ϳ10 Å.

Structure of PSII-LHCII supercomplex
The core heterocomplex (Core) is the highest-resolution part of the structure (Fig. S3, B, D, and F), whereas the minor LHCIIs (CP26 and CP29) and S-trimers are also well-resolved given the sampling frequency used (calibrated to 1.992 Å/pixel at the detector). However, the membrane-extrinsic OEC and M-s and L-trimers are not so well-resolved. PsbP and PsbQ, both components of the OEC that were unstable in the deter- Table 1 Oxygen-evolving activities and Chl a/b ratios of the NAPol-stabilized PSII-LHCII supercomplexes as compared with the ␣-DDM-solubilized supercomplexes Because more Chls are associated with a NAPol-stabilized PSII-LHCII supercomplex, oxygen-evolving activities are compared based on D1 protein amounts, which were quantified by immunodetection.

Structure of C. reinhardtii C 2 S 2 M 2 L 2 -type PSII supercomplex
gent-solubilized supercomplex, are evident in the SDS-PAGE ( Fig. 1A) and can be fit to density in the map (Fig. 2). However, their density is less reliable than the membrane components that are stabilized more by NAPol. The extrinsic densities may be easily distorted by the carbon support film or surface tension effects in thinner regions of ice. This is the best exhibited by PsbQ where density fades quickly at increasing . Using the criteria laid down by Rosenthal and Rubinstein (26) describing the resolutions at which various elements of a cryo-EM map are resolved, the model of 5.8 Å should not be overinterpreted; helices are evident, and some ␤-sheets may be resolved, but side chains are not identifiable (see Movie S1). As shown in Fig. 2D, the 5.8 Å reconstruction of C. reinhardtii PSII-LHCII supercomplex contains three LHCII trimers and two minor LHCs (CP26 and CP29) on both sides of the core dimer. Sixteen membrane intrinsic subunits, PsbA/B/C/D/E/F/ H/I/J/K/L/M/Tc/W/X/Z, and three membrane extrinsic subunits, PsbO/P/Q, were identified. Among the membrane intrinsic subunits, C. reinhardtii possesses a plant-specific PSII subunit, PsbW, which is present in photosynthetic eukaryotes but not in cyanobacteria. PsbW binds on the interface between the core complex and the LHCII array and has been shown to be essential to the formation of PSII-LHCII supercomplex in A. thaliana (27). A schematic representation of the subunit locations is depicted in Fig. 1C.
To evaluate the fitted model, a rigid-body fitting of the coordinates from the spinach PSII-LHCII supercomplex (PDB code 5XNL) was performed. Table S1 contains correlation values for all components. Some individual subunits show higher correlation with the rigid body fit; the correlations of PsbI, PsbJ, and CP29 improved to 0.823, 0.863, and 0.7932, respectively. Many modified tetrapyrrole cofactor densities can be resolved (Fig.  S5, B-F) but not to resolutions required for precise fitting and orientation determination (2).
The overall structure of the purified C. reinhardtii PSII-LHCII complex appears slightly convex when viewed parallel to the membrane (Fig. 2, B and C, and Fig. S4, side views) with the LHCII trimers slightly out-of-plane when compared with the C 2 S 2 M 2 structure from pea of Su et al. (3), although in both C 2 S 2 M 2 structures (3,17) reprojections of the complex appear curved when viewed from certain orientations that would be parallel to the membrane. In others they appear flat. In all

Structure of C. reinhardtii C 2 S 2 M 2 L 2 -type PSII supercomplex
reprojections parallel to the membrane except for one, the C. reinhardtii C 2 S 2 M 2 L 2 PSII-LHCII, they appear curved. In the core regions they are similar; the most significant structural differences are in both the major and minor LHCII regions (Fig.  3) where all LHC complexes are dislocated when compared with those of land plants, whereas the M-trimer has a different angle of rotation ( Fig. 3, C, D, G, and H) relative to the S-trimer when compared with the land plant structures (3,17,18).
The core complex region shows high similarities in general arrangement between C. reinhardtii and the previously published land plant maps, with few major changes beyond the presence of the extra density at the external edge of the complex ( Fig. 1C and Fig. S6). The separate dimers are 4 Å further apart, but each core complex itself is largely structurally unchanged. Of the larger components of the core complex, PsbA (D1), PsbC (CP43), and PsbD (D2) all align well, with PsbB (CP47) also aligning well except for the third helix (at the core complex dimer interface), which is dislocated 2.9Å toward the LHCII array in C. reinhardtii. Of the small-molecular-weight subunits, density associated with PsbX in land plants is turned at an angle of ϳ10°relative to the PsbX in PDB code 3JCU (18) or PDB code 5XNL (3) (Fig. S5G, left panel), whereas PsbH is dislocated 2 Å toward the core protein PsbD in both dimers (Fig.  S5G, right panel). PsbL, M, and Tc all maintain their relative positions in the interface between the two core complexes (Fig.  S5A), as do the edge-associated small subunits PsbE, F, J, and K. C. reinhardtii possesses a PSII complex subunit not present in cyanobacteria, PsbW, which has been shown to be essential to the formation of PSII dimers in land plants (27). Binding on the interface between the core complex and the LHCII array, PsbW shows 3.4 Å dislocation toward the LHCII S-trimer (Fig. S5H), and PsbZ exhibits minor dislocation toward CP26, both of which are further away from the core complex as compared with land plants. PsbTn, the short protein associated with the extrinsic region in spinach (18), was not observed (Fig. S5I). Of the extrinsic OEC proteins, PsbO is evident and appears stably, although with some density weaker at the extreme of the ␤-barrel ( Fig. 2 and Fig. S3), density is present where PsbP is located in the spinach (18) and pea (3) structures but would be difficult to assign because of the density being relatively weak. PsbQ, on the edge of the OEC (Fig. 2 and Fig. S3), shows the parallel ␣-helix character expected from previous reports (3,18); however, the weakness of density implies only partial occupancy of this protein across all particles in the final reconstruction.
CP26 and CP29 are dislocated away from the core complex (Fig. 3) when compared with other PSII-LHCII structures (3,17,18) by up to 15 Å in the case of spinach C 2 S 2 PSII-LHCII. CP29 presents a different orientation in the C 2 S 2 M 2 L 2 form, rotated ϳ10°when compared with C 2 S 2 (18) and ϳ5°when compared with the C 2 S 2 M 2 structures (3, 17) for which the differences may be attributed to the binding of the L-trimer.

A novel region of density
Upon examination of the 2D classes of the C. reinhardtii C 2 S 2 M 2 L 2 PSII-LHCII supercomplex, it was observed that there were two extra points of density ( Fig. 4 and Fig. S6), commonly attributed to transmembrane helices and potentially interacting with PsbJ, PsbK, and PsbZ, that were not present in projection maps of previously reported C 2 S 2 or C 2 S 2 M 2 PSII-LHCII structures in land plants (3, 17, 18) (Fig. S6). This density

Structure of C. reinhardtii C 2 S 2 M 2 L 2 -type PSII supercomplex
was also present in 3D refinements (Figs. 2 and 3). We compared the proteins assigned in the published maps (3,17,18) with the Psb-related genes in the Kyoto Encyclopedia of Genes and Genomes (28) database for respective species and identified Ycf12 and PsbN as not assigned to any point of density in those maps. We putatively assign Ycf12 to this transmembrane density because its counterpart in cyanobacteria has been mapped to the same position in the crystal structure of PSII in Thermosynechococcus elongatus (4). Ycf12 has been shown to be required for optimal PSII function under high light conditions in Synechocystis sp. PCC 6803 (29) and in C. reinhardtii (30). A model of C. reinhardtii Ycf12 was created with I-TASSER (31, 32) (Fig. S7) and docked in place against a 3D refinement of the C 2 S 2 M 2 L 2 form of PSII-LHCII (Fig. 4).

Supramolecular organization of PSII-LHCII supercomplex
The ionic amphipol A8-35-assisted purification strategy for the isolation of the PSII-LHCII supercomplex from C. reinhardtii dramatically improved the yield of the C 2 S 2 M 2 L 2 form of PSII-LHCII supercomplex particles to ϳ56% (11); this is considerably higher than yields obtained with ␣-DDM alone, where the C 2 S 2 M 2 L 2 supercomplex was ϳ11% of the purified protein (10). Here, we utilized a NAPol to further improve the yield of C 2 S 2 M 2 L 2 -type PSII-LHCII supercomplex by replacing the detergent envelope surrounding the supercomplex with NAPol during SDG purification (Fig. 1A): C 2 S 2 M 2 L 2 -type (68% of 'top' views), C 2 S 2 M 1 L 1 -type (31% of 'top' views), and Ͻ1% of the C 2 S 2 -type (Fig. 5), indicating that the major supramolecular organization of the PSII-LHCII supercomplex in C. reinhardtii is the C 2 S 2 M 2 L 2 -type at least under the experimental conditions used in this study (Table 2). These percentages are consistent with the calculations of bound Chl and estimations of total number of bound LHCII trimers and likewise average to ϳ5.4 LHCII trimers per supercomplex.
Because of the heterogeneous nature of PSII-LHCII (10), we were concerned about variation within the biochemical preparation for the different forms as this is not easily measured by SDS-PAGE, immunoblot or oxygen evolution assay. The apparent presence of only three types of PSII-LHCII 'top' views was a curiosity that we had not expected (Fig. 5). To determine whether we could isolate the smaller symmetric particles through using a different reference, we tested multiple picking strategies, the results of which are briefly summarized in Fig. S8. Gautomatch (32) picking software was tested with 50 Å lowpass filtered micrographs and 20 Å low-pass filtered projections of PDB code 3JCU (18), PDB code 5MDX (17) and a PDB code 3JCU-based C 2 S 2 M 2 L 2 -type PSII-LHCII. All projection sets picked roughly the same number of particles, but initial 2D classification varied. Regardless of the picking methodology, the only two strongly evident forms of PSII-LHCII were . Novel transmembrane density alongside the core dimer, which we putatively ascribe to Ycf12 (orange). A, the area containing this previously unreported density is circled in red, 2D reprojection. The contrast has been increased 70%. B, view from the luminal side of PSII-LHCII super complex (the same orientation as the 2D reprojection of A). C, view parallel to the membrane, turned on x and y axes by 90°with NAPol ring erased for clarity. The red arrow in A indicates the direction of view toward the Ycf12 location in C. This novel density appears to coordinate to PsbJ, PsbK, and PsbZ with PsbE and PsbF (pale purple) to the right. CP26 is in mustard yellow. This density is present on both sides of the core complexes for both the C 2 S 2 M 2 L 2 and C 2 S 2 M 1 L 1 PSII-LHCII supercomplexes. Scale bar, 10 nm.

Structure of C. reinhardtii C 2 S 2 M 2 L 2 -type PSII supercomplex
C 2 S 2 M 2 L 2 and C 2 S 2 M 1 L 1 . Combined with the strong low-pass filtering, the highly similar particle selection despite significantly different projections eases concern about missing the smaller symmetric PSII-LHCII particles like C 2 S 2 M 2 .
When the C. reinhardtii C 2 S 2 M 2 L 2 map was compared with the land plants' PSII-LHCII supercomplex models, the core regions aligned well, whereas the LHCII array did not (Fig. 3). The PSII-LHCII supercomplex in C. reinhardtii exhibited greater distances between the core and LHCIIs than those of land plants (Fig. 3, B-D and F-H). The misalignments of the LHCII S-trimer were 15 Å for the C 2 S 2 -type PSII-LHCII from spinach, 6 Å for the C 2 S 2 M 2 -type PSII-LHCII from A. thaliana, and 14 Å for the C 2 S 2 M 2 -type PSII-LHCII from pea, respectively (Fig. 3, F-H, arrows). Another point of interest is that the C 2 S 2 M 2 L 2 -type supercomplex was highly flexible, although its yield was high using NAPol stabilization. This is most evident with M-and L-trimers, which present considerable rotational flexibility (Fig. S9). The rotations of 3.9°for the S-trimer, 18.9°f or the M-trimer and of 26.6°for the L-trimer were observed in the 2D classification images, respectively (Fig. S9, arrows).

Discussion
Using a purification procedure based upon NAPol, we obtained a 5.8 Å cryo-EM map of the C. reinhardtii PSII-LHCII supercomplex containing six LHCII trimers (C 2 S 2 M 2 L 2 ). The NAPol-based purification of PSII-LHCII supercomplex had several advantages. First, population of the complete C 2 S 2 M 2 L 2 supercomplex was increased dramatically. Although it was only 7% among the PSII complexes when the thylakoid membranes were solubilized and kept in ␣-DDM (10), we detected 68% of them when we stabilized the supercomplex with NAPol. Second, the oxygen-evolving activity was preserved. Our recent study indicated that an ionic amphipol A8-35 was also effective to keep peripheral LHCIIs to obtain 56% of C 2 S 2 M 2 L 2 supercomplex (11). However, the oxygen evolving was greatly reduced because of the loss of PsbP and PsbQ, which was rescued by the addition of betaine (11). The NAPol-stabilized PSII-LHCII supercomplex in this study marked even higher oxygen evolving than ␣-DDM-solubilized sample without betaine.
C. reinhardtii is sometimes treated as a "missing link" in the evolution of photosynthesis from cyanobacteria to land plants and shares similarities with both cyanobacteria and land plants, while remaining distinct (6, 8, 34). The presence of PsbW and possibly Ycf12 in the C 2 S 2 M 2 L 2 -type PSII-LHCII supercomplex may reflect this intermediate characteristic. It possesses the largest PSI-LHCI supercomplex reported to date in eukaryotic systems (35,36) while also maintaining the six-trimer-based largest PSII-LHCII supercomplex (10,11,20). The reason why a unicellular alga has developed such large antenna complexes, while higher plants have reduced antenna complexes, is indicative of the low-light environments that C. reinhardtii may find itself exposed to. In lakes or rivers, its exposure to light can be influenced by currents drawing it deeper, wherein only a small quantity of shortwavelength light is able to penetrate, whereas it may be exposed to high-light environments where a large, highly absorbent LHC array is a significant detriment to the overall health of the alga.
When using NAPol-stabilized supercomplexes, purified C. reinhardtii PSII-LHCII does not appear to create supercomplex stacks, as demonstrated for P. sativum (3,19,37), although this behavior has been noted in C. reinhardtii PSII preparations using ␣-DDM (Fig. S10). The pH and Ca 2ϩ dependence of complex stacking (3) indicates the physiological importance for stacking in vivo, which has been demonstrated for thylakoid membranes but not the photosynthetic complexes themselves (19). Because this lack of stacking with NAPol purification is also associated with some possibility that supercomplex stacks exist as purification artifact, future study is needed to clarify this point.
The lack of C 2 S 2 M 2 complex and the presence of a significant proportion of the asymmetric C 2 S 2 M 1 L 1 complex may suggest that, upon dissociation of an L-trimer, the next LHCII trimer lost is the adjoining M-trimer rather than the L-trimer from the opposite side of the complex (Fig. 1C). The results also suggested that the interaction between M-and L-trimers is stronger than that between these trimers and other subunits.
In C. reinhardtii, the PSI-LHCI-LHCII supercomplex can be obtained using growth conditions that favor state 2 conditions (38). However, it was also possible to identify a small proportion of PSI-LHCI-LHCII supercomplexes in normal state 1 conditions present in the PSII-LHCII fraction of an SDG (Fig.  S11). These PSI-LHCI-LHCII supercomplexes can be overlaid with PDB models for the recently reported PSI-LHCI supercomplex (35) and two LHCII trimers (39) (Fig. S9), one of which aligns (albeit at low resolution) to the PSI-LHCI supercomplex with 10 LHCA proteins bound (35,36) (Fig. S11A), whereas the second shows two extra densities of considerable size, which fit well to LHCII trimers (Fig. S11, B and C). There was no apparent class for a single LHCII trimer bound to these PSI species, unlike the PSI-LHCI-LHCII revealed by Pan et al. (40) in maize. It would be an intriguing hypothesis that a pair of M-and L-trimers can be exchanged between PSI-LHCI and PSII-LHCII supercomplexes during state transitions. A hypothetical model of this movement is illustrated in Fig. S11D. To further investigate this problem, we need to achieve close to atomic resolution for multiple species of PSII-LHCII and PSI-LHCI supercomplexes, from both state 1 and state 2 conditions. At the same time, we need to remember the possibility that amphipol may stabilize or even promote the formation of nonnative configurations so that the results obtained in this study need to be verified independently. The structure revealed in Figure 5. Representative 2D projections of the three subspecies of PSII-LHCII supercomplex identified within this study. A, symmetric C 2 S 2 M 2 L 2 type with six LHCII trimers. Particle count is 19,864. B, asymmetric C 2 S 2 M 1 L 1 type with four LHCII trimers. Particle count is 9035. C, symmetric C 2 S 2 type with two LHCII trimers. Particle count is 371. The percentages of the three subspecies stabilized with NAPol are in the bottom right for each particle type. Scale bar, 10 nm.

Structure of C. reinhardtii C 2 S 2 M 2 L 2 -type PSII supercomplex
this study does not reach a sufficiently high resolution to be able to confidently fit cofactors to the density. Moreover, we cannot rule out the possibility of the presence of additional Chls. For these reasons, only a limited energy route mapping was carried out by measuring distances from the center of proteins using UCSF Chimera (Fig. S12). The intermonomer distances for pea PSII-LHCII supercomplex (PDB code 5XNL) (3) are summarized in Fig. S12B, along with the equivalent measurements for the C 2 S 2 M 2 L 2 complex from C. reinhardtii (Fig. S12A). In the C. reinhardtii C 2 S 2 M 2 L 2 -type PSII-LHCII, distances between the core and the minor monomeric LHCIIs are increased, whereas the distance from the S-trimer to the core is increased only slightly. Therefore, excitation transfer from S-trimer may occur more predominantly through the direct pathway to the core in C. reinhardtii (Fig. S12).
At the 2D classification stage, a sharp density was identified that is not present in the published maps of PSII-LHCII in land plants (3,17,18) (Figs. 3-5 and Fig. S6). In the cyanobacterial PSII crystal structure, Umena et al. (4) placed Ycf12 in this location. van Bezouwen et al. (17) discussed weak density in their A. thaliana map in a similar location but did not place any subunit in their model because A. thaliana does not possess Ycf12, and another option, PsbY, is significantly larger than the density. We tentatively assigned this transmembrane density as Ycf12 in this study because C. reinhardtii PSII contains Ycf12 (30). However, we still cannot rule out the possibility that this density might correspond to PsbY, which would be possible if PsbY in C. reinhardtii is expressed as a precursor polyprotein and processed into a few single-transmembrane helix proteins as previously proposed (7).
Another option may be PsbN, because bioinformatics searches revealed PsbN to be a small, 44-residue protein that would fit into density for a single-transmembrane helix, and this location was once assigned to PsbN by Ferreira et al. (41) in their PSII structure in the cyanobacterium T. elongatus. However, this is less likely because the same density was later shown to be Ycf12 by Umena et al. (4) in their atomic resolution PSII structure from cyanobacterium Thermosynechococcus vulcanus. Torabi et al. (42) showed in tobacco that PsbN may play a role in assembly of the PSII complex rather than a subunit in the mature complex. Clarity regarding the identity of this new density will be achieved with a high resolution (ϳ3 Å) cryo-EM map.

Biochemical preparation
Thylakoid membranes from C. reinhardtii cells were prepared as described previously (38). Purification of PSII-LHCII supercomplexes was carried out essentially as previously reported (10) with the following modifications: thylakoid membranes were solubilized with 1.4% ␣-DDM (Anatrace, Maumee, OH) for 10 min in a 25 mM MES buffer (pH 6.5), and unsolubilized membranes were removed by centrifugation at 25,000 ϫ g for 1 min. Subsequently, NAPol (Anatrace) was added at a final concentration of 2% to stabilize the solubilized thylakoids, followed by 10-min dark incubation on ice. The NAPol-stabilized supercomplex was then subjected to ultracentrifugation. For ␣-DDM sample preparation (Fig. 1A, ␣-DDM), ␣-DDM was added to the SDG centrifuge tubes at a concentration of 0.02%. After fractionation of the PSII-LHCII supercomplex band from the SDG, the buffer was changed to 25 mM MES (pH 6.5) on a PD-10 column (GE Healthcare), followed by concentration in a 100-kDa cut-off spin concentrator column (Sartorius, Göttingen, Germany).

SDS-PAGE and immunoblotting
SDS-PAGE was carried out as previously described (45). Amounts of D1 and PsaA proteins were quantified using IMAGE LAB software (Bio-Rad) after immunoblot analysis with antibodies. The antibody used for D1 protein detection (AS05-084) was purchased from Agrisera (Vännäs, Sweden). The antibody used for PsaA detection was described previously (45).

Pigment analysis
Chls were extracted from PSII-LHCII supercomplexes with 80% acetone solution. Chls were separated by ultra-performance LC using a Water H-class system, as previously described (44).

Oxygen-evolving activity
Oxygen-evolving activity was measured at 25°C using a Witrox 4 oxygen meter (Loligo Systems). The isolated PSII-LHCII supercomplexes were suspended at a final concentration of 10 g Chl/ml in 25 mM MES buffer (pH 6.5) containing 5 mM CaCl 2 , 1 mM potassium ferricyanide, and 0.25 mM 2,6-dichlorobenzoquinone. White light at 5,000 mol photons m Ϫ2 s Ϫ1 from a metal halide lamp (Nippon PI Co., Ltd., Tokyo, Japan) was used to activate light-dependent oxygen evolution.

Particle type
This report A8- 35  Other types also found Other types also found Structure of C. reinhardtii C 2 S 2 M 2 L 2 -type PSII supercomplex ence spectra. The ascorbate-reduced minus hydroquinone-oxidized spectrum was used to quantitate the amount of cytochrome b 559 on a Chl basis, which was recorded on a JASCO V-650 spectrophotometer (JASCO Corp., Tokyo, Japan). The extinction coefficient 15 mM Ϫ1 cm Ϫ1 (559 -570 nm) was used (46).

EM data acquisition
An aliquot of purified NAPol-stabilized PSII-LHCII supercomplex (0.25 mg Chl/ml) was applied onto a R 1.2/1.3 Mo grid (Quantifoil Micro Tools) attached an additional thin carbon film for specimen support and treated by glow discharging beforehand. The grid was plunged-frozen using a Vitrobot Mark IV (Thermo Fisher Scientific) and mounted on a 626 cryo-transfer holder (Gatan Inc.). The data acquisition was carried out using a 200kV electron microscope (JEM-2200FS; JEOL Inc.) equipped with an omega-type energy filter and a DE20 direct detector CMOS camera (Direct Electron LP) as previously described (35). The total electron dose for each image was Ͻ30 e Ϫ /Å 2 using a low dose mode. A range of defocus micrographs, totaling 498 micrographs, were manually acquired from 2-to 4-m defocus at a magnification of 30,000ϫ, corresponding to a calibrated resolution of 1.992 Å/pixel on the specimen.

Model fitting
To quantitatively evaluate the 5.8 Å cryo-EM map of the PSII-LHCII supercomplex, a rigid-body fitting of PDB code 5XNL (3) was performed, where the model of S-trimers were first adjusted for strongly binding trimer locations before adjusting the minor LHCs CP26 and CP29; then the model M-and L-trimers were in turn fitted into the densities of the respective positions (Fig. 2,  D-F). This improved the overall correlation of the PSII-LHCII supercomplex from 0.73 to 0.79, although the overall map-tomodel correlation was poor because of large volumes within the map containing unresolved lipids and cofactors and the NAPol ring; cisTEM does not apply more than a spherical mask during final reconstruction (21), so the NAPol density is not masked out.
In attempting to identify the smaller symmetric PSII-LHCII supercomplexes, Gautomatch was used with different refer-ences based upon PDB code 3JCU (18), PDB code 5MDX (17), and a modified PDB code 3JCU with four extra LHCII trimers with and without an extra helix PDB assigned to the novel density. For each of these, references were generated by EMAN2 (33) (e2pdb2mrc module) at a resolution of 20 Å. These resulting maps were then used to create 2D references (e2project3d module) at angular increments of 30°. Micrographs were filtered with a 50 Å low-pass filter. The Gautomatch picks were extracted, and 2D classification was carried out. cisTEM specific procedures Fig. S2 provides a graphical summary of the following methodology. The methods of ab initio reference generation and 3D refinement as carried out in cisTEM are described in detail by Grant et al. (21). After import and CTF fitting of 498 micrographs using the parameters detailed above, cisTEM particle picking was carried out by Gaussian blob, picking a total of 126,744 particles, which were subjected to two rounds of 2D classification to remove contamination, edge-located, densely clustered and/or overlapping particles, and as much macromolecular structural heterogeneity (Figs. S2 and S4) as possible. By default, cisTEM limits the maximum resolution used during 2D classification to 8 Å. Particles were divided into C 2 S 2 M 2 L 2 , C 2 S 2 M 1 L 1 , and side views for further cleaning. Multiple initial models were generated, with the best of the resulting models selected to act as a reference for the autorefinement procedure before rigid-body fitting. Initially, a general spherical mask of 400 Å diameter was imposed in cisTEM, whereas the automasking function was used for initial automatic refinements into two classes, the better of which resolved to 6.16 Å. A low-pass Gaussian filter was applied to the reconstruction using UCSF Chimera, and this resulting map converted into a mask using the cisTEM internal function, "make_size_mask," which was used to mask the reconstruction during local refinement for a final resolution of 5.8 Å (Fig. S2H). The I-TASSER suite (31,32) was used to construct homology models of Ycf12 (Fig. S7) for docking to the empty transmembrane helix density in the final cryo-EM map ( Fig. 4 and Movie S1).